Data and code from: Drivers behind spatiotemporal variation in environmental DNA: An assessment using a rare aquatic salamander, the Eastern Hellbender (<em>Cryptobranchus alleganiensis alleganiensis</em>)

Environmental DNA (eDNA) sampling has become a common tool for monitoring rare and declining aquatic species. However, results can be biased by the spatiotemporal variation in eDNA signals, yet the biological and environmental factors that cause this variation are not well understood. Here, we examined the seasonal and fine-scale (< 100 m) longitudinal variation in eDNA concentration and detection in situ for a rare aquatic salamander, the Eastern Hellbender (Cryptobranchus alleganiensis alleganiensis). We also applied multivariate generalized linear mixed-effects models to investigate how physical and hydrological stream characteristics influence eDNA concentration estimates and detection rates. Both metrics spiked during the hellbender breeding season, but remained low at other times of the year. This was primarily driven by one site in which hellbenders are dense and actively reproducing; once this site was removed from analyses, temporal variation in eDNA signals was no longer observed. No fine-scale spatial eDNA pattern emerged, but concentrations and detections were highly variable across temporal and spatial replicates within sites, emphasizing the importance of collecting replicate eDNA samples at multiple scales. Only PCR inhibitors present in our samples significantly reduced concentrations and detections; however, general negative relationships were still apparent with flow velocity. Increased surface water temperature and pH were also tenuously associated with eDNA concentrations but did not influence detections. Our study contributes important knowledge regarding biological and environmental factors driving eDNA spatiotemporal variability, facilitating a refinement in eDNA sampling strategies for hellbenders and other rare aquatic organisms so that accurate scientific inferences about their populations can be made. Methods Sample Collection We collected water samples from 90 sites across 73 rivers in Kentucky from September 2019 to June 2021. One site located within the Licking River basin was included as a positive field control – this is the only known location in Kentucky where hellbenders are regularly observed and are actively reproducing. Once on location, we visually surveyed for the highest quality habitat available within 500 m upstream and downstream. We defined high-quality habitat as swift running, semi-shallow water located downstream of riffles with large rock slabs, boulders, or bedrock shelving with crevices, and gravel and cobble substrate (Mayasich et al. 2003). A 2 L eDNA sample was collected downstream of the highest quality habitat at all 90 sites during the hellbender breeding season in September and October. Of the 90 eDNA sites, 45 were randomly selected for additional temporal sampling during the November hatching and June pre-breeding seasons. Among the 45 temporally sampled sites, we collected eDNA samples at additional spatial replicates from 33 randomly selected sites. Three spatial replicates were collected: 1) directly downstream of the best available habitat consistent with the protocol used at all 90 sites, 2) 50 m downstream of Replicate 1, and 3) 50 m downstream of Replicate 2. Samples were collected at the most downstream replicate first to prevent potential eDNA contamination caused by upstream sediment. All water samples were collected from the thalweg of the stream or as deep as possible after disturbed sediment was allowed to settle. Two autoclaved 1 L wide-mouth Nalgene bottles with lids attached were placed as close to the bottom of the stream as possible without disturbing the substrate, opened until filled, then resealed while still underwater. Samples were immediately placed on ice after being collected. A 250 mL water sample was then collected using the same technique to estimate conductivity, turbidity, total organic carbon (TOC), and pH. Surface water temperature was measured during each site visit using a digital waterproof thermometer. Flow velocity was also measured during each site visit and at each spatial replicate when appropriate by using a standardized float method (USGS 1982). Floats were conducted three times at each location and averaged for a final flow velocity estimate; if a site had spatial replicates, flow velocity was measured at each replicate and averaged for a single site-level estimate. All sampling equipment was sterilized in 10 % bleach, and clothing was changed in between sites. Negative field controls (i.e., 1 L autoclaved DI water) were used to monitor for cross-contamination between samples each day. Water samples were filtered at the lab within 10 hours after collection through a 0.47 µm mixed cellulose ester membrane (Whatman) by using a sterile Nalgene filter holder with funnel, 1 L Nalgene filter flask, and Rocker 300, 60 Hz vacuum pump (Southern Labware). A negative control consisting of 1 L autoclaved DI water was filtered at the end of each day to test for cross-contamination that may have occurred during filtration. Filters were immediately frozen at -20 °C until eDNA was extracted. All filtering equipment was sterilized in 10% bleach and rinsed in DI water between samples and/or spatial replicates. Laboratory Analysis We extracted DNA from filter membranes following methods described by Spear et al. (2015). Filters were cut into small pieces and extracted using DNeasy Blood and Tissue Kits (Qiagen) following guidelines provided by the manufacturer, with the additional use of a Qiashredder spin column (Qiagen) to facilitate degradation of the filter membrane. Additionally, if multiple filters were needed to complete the filtration of a field sample, we combined these during the binding step of DNA extraction following the protocol of Takahashi et al. (2018). Briefly, the lysate derived from each filter was passed through the same spin column before being washed with AW1 buffer. One blank filter was incorporated into each round of DNA extraction to monitor for cross-contamination during DNA extraction. We amplified eDNA samples at a 104 bp sequence of the mitochondrial cytochrome b region using nine quantitative PCR (qPCR) replicates per sample with primers and probes designed by Spear et al. (2015; Table S1 in associated manuscript). We ran 15 µL qPCR reactions on a QuantStudio3 thermocycler (Applied Biosystems) with 7.5 µL TaqMan Environmental Master Mix 2.0 (ThermoFisher Scientific Inc.), 0.6 µL (0.4 µM concentration) of each primer, 0.3 µL (0.2 µM concentration) of probe, 0.6 µL TaqMan Exogenous Internal Positive Control 10X Exo IPC Mix (Applied Biosystems), 0.3 µL TaqMan Exogenous Internal Positive Control 50X Exo IPC DNA (Applied Biosystems), 0.12 µL (0.4 µg/µL concentration) BSA, 1.98 µL nuclease-free water, and 3 µL of eDNA template. TaqMan Exogenous Internal Positive Controls were added to each sample reaction to monitor for inhibition. Thermo-cycling parameters followed those recommended for use with TaqMan Environmental Master Mix 2.0 and began with 10 min at 95 °C, followed by 45 cycles of 95 °C for 15 s and 60 °C for 1 min. Each plate contained three positive template controls obtained by extracting eDNA from aquarium tank water housing adult hellbenders and three negative template controls consisting of DI water. Triplicates of a standard dilution series composed of five or six 10x-fold dilutions ranging from 9.4-9.4x10 -5ng/µL or 1.14-1.14x10 -5 ng/µL were also included on each plate to estimate eDNA concentrations. Two hellbender tissue samples (94.0 and 11.4 ng/µL) were needed to develop our standard curves because all of the extracted DNA from our first sample was used in this study. DNA extraction, qPCR prep, and qPCR were conducted in separate rooms dedicated to each task to reduce the risk of cross-contamination. We used QuantStudio Design and Analysis Software v1.5.1 (ThermoFisher Scientific Inc) to estimate eDNA concentrations in each qPCR replicate and calculate detection rates for each temporal or spatial replicate. We considered a sample positive when at least one of the nine qPCR replicates amplified, the curve morphology was uniform with the standard replicates, and the negative qPCR controls showed no amplification (Klymus et al. 2020).